search

kjunelee / MetaOptNet

Meta-Learning with Differentiable Convex Optimization (CVPR 2019 Oral)
Apache License 2.0
529 stars 97 forks source link

Why dual formulation in Bertinetto's regression solver? #36

Closed arxrean closed 4 years ago

arxrean commented 4 years ago

Hi, I'm new to this. I read both your paper and Bertinetto's. In their paper, the closed form solver is derived by classicial regression or Newton’s method. There is no dual form needed in theirs. But in your paper, their solution has a dual form. Does it contain some hidden processes? Thank you!

kjunelee commented 4 years ago

You can find the derivation of dual ridge regression e.g. in "Ridge Regression Learning Algorithm in Dual Variables" by Saunders et al. (1998).

Hope this helps.